MahdyTech

An infrequent issue on many machines can disguise as a cluster-wide issue.

It hit me that I don’t know what Large Pages really are, and I needed deeper info - which I am going through in this post.

VTune is an instruction level profiler. I want to explore if it can detect cache unfriendliness, and fix it.

Latency was high at low QPS due to an OS optimization. The profile did not catch the delays because they were on the Kernel side of things.

UMDH is a tightly stripped version of XPerf's extended leaks debugging capabilities. While not nearly as powerful, in several scenarios it is good enough.

Cooperatively-scheduling our threads to yield on IO blocks allow us to process much more requests per second, with no reprimands for latency.

Heap corruptions can be nasty to investigate. I rebuilt in debug & voila! The app was throwing exactly where the problem was happening.

Spawning a RESTful C++ app on an Azure fleet, using C++ REST SDK for building the web server and Service Fabric for orchestration.

Latency hike was caused by something local. However, `Lucy` had no new deployments for some time, no new data, nothing! How can nothing trigger something?

Task Manager can be lacking as a memory tracker. Let's cover alternatives to replace it. First, let’s discuss how memory allocations work in Windows.

I used to think shutdowns are easy. Unfortunately that’s not completely the case. Ungraceful shutdowns bring about crashes, or unpredicted change to state.

I used to think shutdowns are easy. Unfortunately that’s not completely the case. Ungraceful shutdowns bring about crashes, or unpredicted change to state.